hclustercommand.cpp

   1 /*
   2  *  hclustercommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 10/13/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "hclustercommand.h"
  11
  12 //**********************************************************************************************************************
  13 //This function checks to make sure the cluster command has no errors and then clusters based on the method chosen.
  14 HClusterCommand::HClusterCommand(string option){
  15         try{
  16                 globaldata = GlobalData::getInstance();
  17                 abort = false;
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string Array[] =  {"cutoff","precision","method","phylip","column","name","sorted","showabund","timing"};
  25                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  26
  27                         OptionParser parser(option);
  28                         map<string,string> parameters = parser.getParameters();
  29
  30                         ValidParameters validParameter;
  31
  32                         //check to make sure all parameters are valid for command
  33                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  34                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {
  35                                         abort = true;
  36                                 }
  37                         }
  38
  39                         globaldata->newRead();
  40
  41                         //check for required parameters
  42                         phylipfile = validParameter.validFile(parameters, "phylip", true);
  43                         if (phylipfile == "not open") { abort = true; }
  44                         else if (phylipfile == "not found") { phylipfile = ""; }
  45                         else {  distfile = phylipfile;  format = "phylip";      }
  46
  47                         columnfile = validParameter.validFile(parameters, "column", true);
  48                         if (columnfile == "not open") { abort = true; }
  49                         else if (columnfile == "not found") { columnfile = ""; }
  50                         else {  distfile = columnfile; format = "column";       }
  51
  52                         namefile = validParameter.validFile(parameters, "name", true);
  53                         if (namefile == "not open") { abort = true; }
  54                         else if (namefile == "not found") { namefile = ""; }
  55
  56                         if ((phylipfile == "") && (columnfile == "")) { mothurOut("When executing a hcluster command you must enter a phylip or a column."); mothurOutEndLine(); abort = true; }
  57                         else if ((phylipfile != "") && (columnfile != "")) { mothurOut("When executing a hcluster command you must enter ONLY ONE of the following: phylip or column."); mothurOutEndLine(); abort = true; }
  58
  59                         if (columnfile != "") {
  60                                 if (namefile == "") {  cout << "You need to provide a namefile if you are going to use the column format." << endl; abort = true; }
  61                         }
  62
  63                         //check for optional parameter and set defaults
  64                         // ...at some point should added some additional type checking...
  65                         //get user cutoff and precision or use defaults
  66                         string temp;
  67                         temp = validParameter.validFile(parameters, "precision", false);
  68                         if (temp == "not found") { temp = "100"; }
  69                         //saves precision legnth for formatting below
  70                         length = temp.length();
  71                         convert(temp, precision);
  72
  73                         temp = validParameter.validFile(parameters, "cutoff", false);
  74                         if (temp == "not found") { temp = "10"; }
  75                         convert(temp, cutoff);
  76                         cutoff += (5 / (precision * 10.0));
  77
  78                         method = validParameter.validFile(parameters, "method", false);
  79                         if (method == "not found") { method = "nearest"; }
  80
  81                         if ((method == "furthest") || (method == "nearest") || (method == "average")) { }
  82                         else { mothurOut("Not a valid clustering method.  Valid clustering algorithms are furthest, nearest or average."); mothurOutEndLine(); abort = true; }
  83
  84                         showabund = validParameter.validFile(parameters, "showabund", false);
  85                         if (showabund == "not found") { showabund = "T"; }
  86
  87                         sort = validParameter.validFile(parameters, "sorted", false);
  88                         if (sort == "not found") { sort = "F"; }
  89                         sorted = isTrue(sort);
  90
  91                         timing = validParameter.validFile(parameters, "timing", false);
  92                         if (timing == "not found") { timing = "F"; }
  93
  94
  95                         if (abort == false) {
  96
  97                                 fileroot = getRootName(distfile);
  98
  99                                 tag = "fn";  //until we figure out average and nearest methods
 100
 101                                 openOutputFile(fileroot+ tag + ".sabund",       sabundFile);
 102                                 openOutputFile(fileroot+ tag + ".rabund",       rabundFile);
 103                                 openOutputFile(fileroot+ tag + ".list",         listFile);
 104                         }
 105                 }
 106         }
 107         catch(exception& e) {
 108                 errorOut(e, "HClusterCommand", "HClusterCommand");
 109                 exit(1);
 110         }
 111 }
 112
 113 //**********************************************************************************************************************
 114
 115 void HClusterCommand::help(){
 116         try {
 117                 mothurOut("The hcluster command parameter options are cutoff, precision, method, phylip, column, name, showabund, timing and sorted. Phylip or column and name are required.\n");
 118                 mothurOut("The phylip and column parameter allow you to enter your distance file, and sorted indicates whether your column distance file is already sorted. \n");
 119                 mothurOut("The name parameter allows you to enter your name file and is required if your distance file is in column format. \n");
 120                 mothurOut("The hcluster command should be in the following format: \n");
 121                 mothurOut("hcluster(column=youDistanceFile, name=yourNameFile, method=yourMethod, cutoff=yourCutoff, precision=yourPrecision) \n");
 122                 mothurOut("The acceptable hcluster methods is furthest, but we hope to add nearest and average in the future.\n\n");
 123         }
 124         catch(exception& e) {
 125                 errorOut(e, "HClusterCommand", "help");
 126                 exit(1);
 127         }
 128 }
 129
 130 //**********************************************************************************************************************
 131
 132 HClusterCommand::~HClusterCommand(){}
 133
 134 //**********************************************************************************************************************
 135
 136 int HClusterCommand::execute(){
 137         try {
 138
 139                 if (abort == true) {    return 0;       }
 140
 141                 if(namefile != ""){
 142                         globaldata->nameMap = new NameAssignment(namefile);
 143                         globaldata->nameMap->readMap();
 144                 }else{
 145                         globaldata->nameMap = NULL;
 146                 }
 147
 148                 time_t estart = time(NULL);
 149
 150                 if (!sorted) {
 151                         read = new ReadCluster(distfile, cutoff);
 152                         read->setFormat(format);
 153                         read->read(globaldata->nameMap);
 154                         distfile = read->getOutputFile();
 155
 156                         list = read->getListVector();
 157                         delete read;
 158                 }else {
 159                         list = new ListVector(globaldata->nameMap->getListVector());
 160                 }
 161
 162                 mothurOut("It took " + toString(time(NULL) - estart) + " seconds to sort. "); mothurOutEndLine();
 163                 estart = time(NULL);
 164
 165                 //list vector made by read contains all sequence names
 166                 if(list != NULL){
 167                         rabund = new RAbundVector(list->getRAbundVector());
 168                 }else{
 169                         mothurOut("Error: no list vector!"); mothurOutEndLine(); return 0;
 170                 }
 171
 172                 float previousDist = 0.00000;
 173                 float rndPreviousDist = 0.00000;
 174                 oldRAbund = *rabund;
 175                 oldList = *list;
 176
 177                 print_start = true;
 178                 start = time(NULL);
 179
 180                 ifstream in;
 181                 openInputFile(distfile, in);
 182                 string firstName, secondName;
 183                 float distance;
 184
 185                 cluster = new HCluster(rabund, list);
 186                 vector<seqDist> seqs; seqs.resize(1); // to start loop
 187                 exitedBreak = false;  //lets you know if there is a distance stored in next
 188
 189                 while (seqs.size() != 0){
 190
 191                         seqs = getSeqs(in);
 192                         random_shuffle(seqs.begin(), seqs.end());
 193
 194                         if (seqs.size() == 0) { break; } //there are no more distances
 195
 196                         for (int i = 0; i < seqs.size(); i++) {  //-1 means skip me
 197
 198                                 if (print_start && isTrue(timing)) {
 199                                         mothurOut("Clustering (" + tag + ") dist " + toString(distance) + "/"
 200                                                           + toString(roundDist(distance, precision))
 201                                                           + "\t(precision: " + toString(precision) + ")");
 202                                         cout.flush();
 203                                         print_start = false;
 204                                 }
 205
 206         //cout << "before cluster update" << endl;
 207                                 if (seqs[i].seq1 != seqs[i].seq2) {
 208                                         cluster->update(seqs[i].seq1, seqs[i].seq2, seqs[i].dist);
 209
 210                                         float rndDist = roundDist(seqs[i].dist, precision);
 211                 //cout << "after cluster update clusterSomething = " << clusteredSomething << " rndDist = " << rndDist << " rndPreviousDist = " << rndPreviousDist << endl;
 212
 213
 214                                         if((previousDist <= 0.0000) && (seqs[i].dist != previousDist)){
 215                                                 printData("unique");
 216                                         }
 217                                         else if((rndDist != rndPreviousDist)){
 218                                                 printData(toString(rndPreviousDist,  length-1));
 219                                         }
 220
 221                                         previousDist = seqs[i].dist;
 222                                         rndPreviousDist = rndDist;
 223                                         oldRAbund = *rabund;
 224                                         oldList = *list;
 225                                 }
 226                         }
 227                 }
 228
 229                 in.close();
 230
 231                 if (print_start && isTrue(timing)) {
 232                         //mothurOut("Clustering (" + tag + ") for distance " + toString(previousDist) + "/" + toString(rndPreviousDist)
 233                                          //+ "\t(precision: " + toString(precision) + ", Nodes: " + toString(matrix->getNNodes()) + ")");
 234                         cout.flush();
 235                         print_start = false;
 236                 }
 237
 238                 if(previousDist <= 0.0000){
 239                         printData("unique");
 240                 }
 241                 else if(rndPreviousDist<cutoff){
 242                         printData(toString(rndPreviousDist, length-1));
 243                 }
 244
 245                 //delete globaldata's copy of the sparsematrix and listvector to free up memory
 246                 delete globaldata->gListVector;  globaldata->gListVector = NULL;
 247
 248                 //saves .list file so you can do the collect, rarefaction and summary commands without doing a read.list
 249                 if (globaldata->getFormat() == "phylip") { globaldata->setPhylipFile(""); }
 250                 else if (globaldata->getFormat() == "column") { globaldata->setColumnFile(""); }
 251
 252                 globaldata->setListFile(fileroot+ tag + ".list");
 253                 globaldata->setNameFile("");
 254                 globaldata->setFormat("list");
 255
 256                 sabundFile.close();
 257                 rabundFile.close();
 258                 listFile.close();
 259
 260                 delete cluster;
 261                 //if (isTrue(timing)) {
 262                         mothurOut("It took " + toString(time(NULL) - estart) + " seconds to cluster. "); mothurOutEndLine();
 263                 //}
 264                 return 0;
 265         }
 266         catch(exception& e) {
 267                 errorOut(e, "HClusterCommand", "execute");
 268                 exit(1);
 269         }
 270 }
 271
 272 //**********************************************************************************************************************
 273
 274 void HClusterCommand::printData(string label){
 275         try {
 276                 if (isTrue(timing)) {
 277                         mothurOut("\tTime: " + toString(time(NULL) - start) + "\tsecs for " + toString(oldRAbund.getNumBins())
 278                      + "\tclusters. Updates: " + toString(loops)); mothurOutEndLine();
 279                 }
 280                 print_start = true;
 281                 loops = 0;
 282                 start = time(NULL);
 283
 284                 oldRAbund.setLabel(label);
 285                 if (isTrue(showabund)) {
 286                         oldRAbund.getSAbundVector().print(cout);
 287                 }
 288                 oldRAbund.print(rabundFile);
 289                 oldRAbund.getSAbundVector().print(sabundFile);
 290
 291                 oldList.setLabel(label);
 292                 oldList.print(listFile);
 293         }
 294         catch(exception& e) {
 295                 errorOut(e, "HClusterCommand", "printData");
 296                 exit(1);
 297         }
 298
 299
 300 }
 301 //**********************************************************************************************************************
 302 vector<seqDist> HClusterCommand::getSeqs(ifstream& filehandle){
 303         try {
 304                 string firstName, secondName;
 305                 float distance, prevDistance;
 306                 vector<seqDist> sameSeqs;
 307                 prevDistance = -1;
 308
 309                 //if you are not at the beginning of the file
 310                 if (exitedBreak) {
 311                         sameSeqs.push_back(next);
 312                         prevDistance = next.dist;
 313                         exitedBreak = false;
 314                 }
 315
 316                 //get entry
 317                 while (filehandle) {
 318
 319                         filehandle >> firstName >> secondName >> distance;
 320 //cout << firstName << '\t' << secondName << '\t' << distance << endl;
 321                         gobble(filehandle);
 322
 323                         //save first one
 324                         if (prevDistance == -1) { prevDistance = distance; }
 325         //cout << prevDistance << endl;
 326 //if (globaldata->nameMap == NULL) { cout << "null" << endl; }
 327                         map<string,int>::iterator itA = globaldata->nameMap->find(firstName);
 328                         map<string,int>::iterator itB = globaldata->nameMap->find(secondName);
 329
 330                         if(itA == globaldata->nameMap->end()){
 331                                 cerr << "AAError: Sequence '" << firstName << "' was not found in the names file, please correct\n"; exit(1);
 332                         }
 333                         if(itB == globaldata->nameMap->end()){
 334                                 cerr << "ABError: Sequence '" << secondName << "' was not found in the names file, please correct\n"; exit(1);
 335                         }
 336         //cout << "here" << endl;
 337                         //using cutoff
 338                         if (distance > cutoff) { break; }
 339
 340                         if (distance != -1) { //-1 means skip me
 341
 342                                 //are the distances the same
 343                                 if (distance == prevDistance) { //save in vector
 344                                         seqDist temp;
 345                                         temp.seq1 = itA->second;
 346                                         temp.seq2 = itB->second;
 347                                         temp.dist = distance;
 348                                         sameSeqs.push_back(temp);
 349                                         exitedBreak = false;
 350                                         //what about precision??
 351
 352                                 }else{
 353                                         next.seq1 = itA->second;
 354                                         next.seq2 = itB->second;
 355                                         next.dist = distance;
 356                                         exitedBreak = true;
 357                                         break;
 358                                 }
 359
 360                         }
 361                 }
 362
 363                 return sameSeqs;
 364         }
 365         catch(exception& e) {
 366                 errorOut(e, "HClusterCommand", "getSeqs");
 367                 exit(1);
 368         }
 369
 370
 371 }
 372
 373 //**********************************************************************************************************************
 374